mstorsjo added a reviewer: rnk.
mstorsjo added a subscriber: rnk.
mstorsjo added a comment.
So, using `-l:libunwind.dll.a` is definitely more correct than
`-l:libunwind.so` on this platform, so in that sense this is good.
In my toolchains I don't use this option at all so far, because I have
libunwind bundled into libc++.a/libc++.dll.a, but I guess it would be nice to
move towards this setup and stop bundling them. But the toolchains can be
configured to omit either the shared or the static version of the libraries,
and if libunwind only exists in static form, this would fail.
If there's only a shared version, and building with `-static-libgcc`, I guess
it's ok to look for explicitly `libunwind.a` and fail in that case, but for the
default case, I think the normal library resolution logic would be best, i.e.
just doing `-lunwind`.
The ideal version from my point of view would be this:
case ToolChain::UNW_CompilerRT:
if (LGT == LibGccType::StaticLibGcc)
CmdArgs.push_back("-l:libunwind.a");
else if (TC.getTriple().isOSCygMing()) {
if (LGT == LibGccType::SharedLibGcc)
CmdArgs.push_back("-l:libunwind.dll.a");
else
CmdArgs.push_back("-lunwind"); // Let the linker choose between
libunwind.dll.a and libunwind.a depending on what's available, and depending on
the -static flag
} else
CmdArgs.push_back("-l:libunwind.so");
break;
That's rather different than the other, non-mingw cases though, but I'm not
familiar with the reasoning behind those.
Adding @rnk if he'd happen to have opinions.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D79995/new/
https://reviews.llvm.org/D79995
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
